◆ Stacks
MISS THE CONSENSUS?
Jukan (@jukan05) · 2026-07-28 · original: EN

Why did SK Hynix's operating profit miss consensus, even with pretax profit landing above ₩100 trillion just as estimated three months ago?

Open in the Stacks app → Read the original ↗

SK hynix (Market Cap: KRW 1,132.2632 trillion) 2Q26 Earnings Results

Revenue: KRW 79 trillion (+257% YoY) (Consensus: KRW 84 trillion; 5.5% below consensus)

Operating profit: KRW 61 trillion (+557% YoY) (Consensus OP: KRW 64 trillion; 5.4% below consensus)

Jukan (@jukan05) · 2026.07.28

Three months ago, Meritz Securities put out an estimate. SK Hynix's Q2 pretax profit would top 100 trillion won, a first for any KOSPI company. The catch: more than 40 trillion won of that wouldn't come from chip sales, but from an eight-year-old Kioxia stake.

What SK Hynix actually sells is DRAM and HBM modules like this one. More than half of Q2's result came from that core business.
What SK Hynix actually sells is DRAM and HBM modules like this one. More than half of Q2's result came from that core business. · Wikimedia Commons

On the 28th, the actual scorecard landed.

How close was the estimate

Revenue came in at 79.3187 trillion won, operating profit at 60.5426 trillion won. Up 257% and 557% year over year. That operating profit number nearly matches the 60.1 trillion won Meritz flagged three months ago. The core chip business moved roughly as expected.

Revenue (actual)₩79.32tn
Operating profit (actual)₩60.54tn
Operating profit (consensus)₩64tn

Both revenue and operating profit missed the market consensus by about 5%. The chip business itself tracked Meritz's estimate, but it still fell short of what Wall Street and Yeouido analysts were pricing in.

2026-07-28, SK Hynix earnings release (PR Newswire)

So why the consensus miss

That doesn't quite add up on the surface. Why would the top AI-memory maker miss expectations at all. Reuters reporting points the other way. SK Hynix carries an unusually heavy mix of premium AI data-centre memory, chips like HBM. This quarter, though, it was ordinary DRAM and NAND pricing that jumped harder than expected. Rivals with more commodity-memory exposure captured more of that upside.

SK Hynix posts sixfold rise in Q2 profit on AI chip demand, misses forecasts (Reuters via Investing.com, 2026-07-28) ↗
The bearish read

The AI-memory leader still missing consensus signals the chip upcycle is cooler than it looks

VS
The one-off read

This was a quarter when commodity memory alone spiked, and SK Hynix's heavier AI-product mix meant it captured relatively less of that specific pop

Same miss, different conclusions depending on which cycle you use as the baseline.

Net income still dwarfs operating profit

Net income came in at 93.9226 trillion won, well above the 60.5426 trillion won operating profit. That's a 118% net margin. The gap is the same Kioxia stake flagged in the earlier estimate. Money SK Hynix put in alongside Bain Capital back in 2018, buying into what was then Toshiba Memory, got cashed out as Kioxia's value climbed after its own listing. The direction matches Meritz's three-month-old estimate. What's different is that these are now the company's own confirmed numbers.

What do the 10 customer deals change

SK Hynix said it began mass-production shipments of HBM4 this quarter, with output set to expand in the second half. At the same time, it completed long-term supply agreements with roughly 10 major customers, deals that lock in volume and price years ahead. Korea Herald ties this to rising demand for AI agents that handle complex, autonomous tasks. As AI moves past chatbots into systems that act on their own, more customers want years of supply locked in.

AI demand drives SK hynix's preliminary Q2 operating profit to 60.5T won (Korea Herald, 2026-07-29)
출처: koreaherald.com

So what settles it

The question comes down to one thing. Which one previews next quarter: the 93 trillion won net profit, or the operating miss. If operating profit is the true scorecard for the core chip business, watch whether the consensus miss repeats next quarter. If this miss was a one-off tied to an unusually sharp jump in commodity memory prices, that gap should narrow. The Kioxia windfall is now mostly cashed out. From next quarter on, the operating profit line alone will show the chip cycle, without anything else mixed in.

In three lines

What happened next

Awaiting gradingTo be scored on whether Q3 2026 operating profit (due late October) again misses its consensus, and whether the gap narrows from this quarter's roughly 5%.

This author's record

51posts27directional calls22Bull5Bear
See the full record →